Poly(γ-glutamic acid) in Aqueous Solution: Molecular Dynamics

David Zanuy, and Carlos Alemán* .... The trajectory of the fraction of i to i + 3 hydrogen bonds, which is relative to 18 and 9 initial hydrogen bond...
2 downloads 0 Views 167KB Size
Biomacromolecules 2001, 2, 651-657

651

Poly(γ-glutamic acid) in Aqueous Solution: Molecular Dynamics Simulations of 10- and 20-Residue Chains at Different Temperatures David Zanuy and Carlos Alema´ n* Departament d’Enginyerı´a Quı´mica, E.T.S.E.I.B., Universitat Polite` cnica de Catalunya, Diagonal 647, Barcelona E-08028, Spain Received November 13, 2000

A series of molecular dynamics simulations in aqueous solution have been carried out in order to investigate the effects of the length of the peptide chain and the temperature on the helical conformation of un-ionized poly(γ-D-glutamic acid). The results reveal that the helix is not stable for a small number of residues independent of the temperature. On the other hand, the temperature induces a conformational transition from the helical state to the random coil. Analysis of the energy of the whole system indicates that the helix is more stable than the random coil state by about 5 kcal/mol, even although the solvation potential energy is lower in the latter than in the former. Introduction Poly(γ-glutamic acid), abbreviated PGGA, is an exopolymer naturally produced by a variety of species of genus Bacillus, like Bacillus lichenformis and Bacillus subtilis.1 This poly(γ-amino acid) has been extensively investigated in terms of its synthesis, biochemistry, and production under different environmental conditions.2-6

PGGA can adopt various conformations depending on solution conditions. Thus, as for most charged or polar biological macromolecules, factors like temperature, pH, polymer concentration, molecular weight, and ionic strength can change the overall conformation and local structure of PGGA. The influence of these factors has been investigated by a variety of techniques. Results revealed that in aqueous solution at low pH, polymer concentration, and ionic strength, the un-ionized PGGA adopts a stable helical arrangement stabilized by intramolecular hydrogen bonds.7,8 The helical conformation of un-ionized poly(γ-D-glutamic acid), denoted as PGDGA, was recently investigated by atomic-resolution computer simulations.9.10 It was found that PGDGA forms a left-handed helix with 19-membered ring hydrogen bonds set between the CO of the amide group i and the NH of the amide group i + 3 (Figure 1), this conformation being fully compatible with the optical rotatory dispersion experiments.9 Furthermore, this helix was predicted to be particularly stable in aqueous solution.10 * Corresponding author. E-mail: [email protected].

In this work, we investigate the influence of both the length of the polymer chain and the temperature on stability of the helical conformation of PGDGA by using molecular dynamics (MD) simulations. For this purpose, MD simulations of 20- and 10-residue chains of un-ionized PGDGA solvated by explicit water molecules were performed at two different temperatures, 300 and 350 K. The outline of the present paper is as follows. In the next section, we briefly describe the theoretical methods used in MD simulations. After this, the results of the four MD simulations are compared. The influence of both the temperature and length of the peptide chain on the dynamics of the helical conformation of PGDGA is discussed. Furthermore, the helix and the random coil states are energetically compared in terms of peptidepeptide (p-p), peptide-water (p-w), and water-water (ww) interactions. Theoretical Methods The systems simulated consisted of 20- and 10-residue PGDGA chains, which were denoted as PGDGA20 and PGDGA10, respectively, and the surrounding water molecules. The all-atom parametrization of the Amber force field11 was used for the PGDGA chains, of which the N-terminus and the C-terminus were acetylated and methylamidated, respectively. Electrostatic parameters for the γ-Dglutamic acid residue, which are not included in the Amber libraries, were taken from a previous work.9 The TIP3P water12 was used as the water model. PGDGA20 and PGDGA10 chains were immersed into a rectangular box of water molecules having a density 1 g/cm3, taken from a Monte Carlo equilibrium simulation. The box was truncated so that there were at least an 8 Å distance between any peptide atom and the edge of the box, and water molecules that overlapped the peptide were discarded. The resulting systems for PGDGA20 and PGDGA10 had dimensions of 45.0

10.1021/bm000122j CCC: $20.00 © 2001 American Chemical Society Published on Web 06/23/2001

652

Biomacromolecules, Vol. 2, No. 3, 2001

Zanuy and Alema´ n

Figure 2. Backbone atom-positional root-mean square deviation (rmsd) from the model conformation of PGDGA as a function of simulation time for PGDGA20 (a) and PGDGA10 (b) at 300 (solid line) and 350 K (dashed line).

Figure 1. Equatorial and axial projections of the model conformation of PGDGA.

× 23.7 × 24.6 and 35.7 × 20.4 × 20.2 Å3, respectively, and contained 731 and 422 water molecules. Periodic boundary conditions were applied using the nearest image convention. The initial conformation for the PGDGA20 and PGDGA10 chains was that proposed on the basis of optical rotatory dispersion experiments and molecular modeling studies.7,9 This model conformation, which is displayed in Figure 1, consists of a left-handed helix stabilized by intramolecular hydrogen bonds and the following torsional parameters: φ ) 71.0°, ξ1 ) 53.0°, ξ2 ) -171.6°, ψ ) 160.6°, and χ1 ) 155.0°. MD simulations of PGDGA20 and PGDGA10 were done at 1 atm of pressure and at two different temperatures, 300 and 350 K. We first performed 2000-step energy minimizations of the complete systems, i.e., the initial boxes containing the solvent and the PGDGA chains. These were followed by 25 ps of NVT-MD (constant volume and temperature MD) during which the temperature was brought to the desired values. During this initial equilibration, the peptides were

restrained to the position resulting from the energy minimization with Cartesian harmonic restraints. Then, the systems were equilibrated for 75 ps of NVT-MD during which both the solute and the solvent were allowed to move freely. After that, we conducted four simulations, PGDGA20 and PGDGA10 at 300 and 350 K, consisting of a total of 1000 ps for each one. Finally, the simulation of PGDGA20 at 300 K was extended to 1500 ps in order to ensure the stability of the helix conformation. Residue-based cutoffs were applied at 10 Å; i.e., if two residues or a residue and a water molecule have any atoms within 10 Å, the interaction between the entire pair is evaluated. The SHAKE algorithm13 was applied to fix the bond lengths, and a numerical integration time step of 2 fs was used for all the simulations. The nonbonded pair list was updated every 25 steps. The MD trajectories were saved every 500 steps (1 ps interval) for subsequent analysis. All calculations were performed with Amber 4.1. computer program14 on a SGI indigo2 system with a single R10000 processor at our laboratory. Results and Discussion Helical Conformation of PGDGA. To ascertain the differences between the model helix used as starting point in MD simulations (Figure 1) and the structures sequentially stored along the trajectory, the atom-positional root-meansquare deviations (rmsd) and both the amount and position of the hydrogen bonds set between the CO of amide i and the NH of amide i + 3 were monitored. Figure 2 shows the rmsd for the backbone atoms (N, CR, Cβ, Cγ, and C) of PGDGA20 and PGDGA10 from the model helix as a function of simulation time. The trajectory of the fraction of i to i + 3 hydrogen bonds, which is relative to 18 and 9 initial hydrogen bonds possible for PGDGA20 and PGDGA10,

Poly(γ-glutamic acid) in Aqueous Solution

Figure 3. Trajectory of the fraction of i to i + 3 type hydrogen bonds in PGDGA20 (a) and PGDGA10 (b) at 300 (solid line) and 350 K (dashed line). An hydrogen bond was considered as formed when the O‚‚‚H distance was less than 2.5 Å and the ∠N-H‚‚‚O angle was larger than 135°.

respectively, and the spatiotemporal evolution of the hydrogen bonds are shown in Figures 3 and 4. The criteria used to define a hydrogen bond were that the O‚‚‚H distance was less than 2.5 Å and the ∠N-H‚‚‚O angle was larger than 135°. Furthermore, we investigated the possible existence of other types of intrapeptide hydrogen bonds between amide groups close in the space. More specifically, we examined the hydrogen bonds set between the CO of amide group i

Biomacromolecules, Vol. 2, No. 3, 2001 653

and the NH of amide groups i + 1, i + 2, and i + 4 as well as those between the NH of amide group i and the CO of amide groups i + 1, i + 2, i + 3, and i + 4. In all cases these types of hydrogen bonds were rarely detected. Thus, for a given snapshot the maximum numbers of intrapeptide hydrogen bonds belonging to a type other than that of the model helix were three and two for PGDGA20 and PGDGA10, respectively, all of them being short-lived. It is worth noting that in the MD simulation of PGDGA20 at 300 K the fluctuations of the helix are small (Figure 2a), the average rmsd along the trajectory being 2.99 ( 1.08 Å. This value decreases to 1.87 ( 1.16 Å for the atoms involved in the 15 central residues. The central part of the helix presents very small fluctuations whereas the fluctuations are larger in the two tails. This is consistent with the evolution of the i to i + 3 hydrogen bonds along the trajectory. Thus, 14 initial hydrogen bonds remain practically intact after 700 ps of simulation, and more than 9 are retained at the end of the simulation (Figure 3a). Figure 4a reveals that the broken hydrogen bonds are essentially located at the tails of the helix, especially on the N-terminus side. On the other hand, results displayed in Figures 2, 3, and 4 indicate that the helix remain stable in the last 500 ps of simulation. The unfolding of the tails is represented in Figure 5a, which shows an atomistic picture of the structures extracted at times 200, 400, 600, 800, and 1000 ps. On the other hand, Figure 6 shows the evolution of the backbone dihedral angles φ, ξ1, ξ2, and ψ along the simulation for the central residue of PGDGA20, in which the helix conformation is retained. The conformation of such angles is gauche+, gauche+, trans, and trans, respectively, indicating that helical fold of PGDGA is induced by the gauche+-gauche+ sequence of the dihedral angles φ and

Figure 4. Spatiotemporal evolution of i to i + 3 type hydrogen bonds: (a) PGDGA20 at 300 K, (b) PGDGA20 at 350 K, (c) PGDGA10 at 300 K, and (d) PGDGA10 at 350 K. Note that the residue number, i, denotes the ith residue with the acceptor oxygen.

654

Biomacromolecules, Vol. 2, No. 3, 2001

Zanuy and Alema´ n

Figure 5. Ribbon snapshots representing equally spaced intervals over the 1000 ps of the PGDGA20 simulations at 300 (a) and 350 K (b). The N-terminus is at the top of the figure.

Figure 6. Evolution of the backbone dihedral angles for the central residue of PGDGA20 through the MD simulation at 300 K.

ξ1. This sequence is also present in the R-helix of poly(Rpeptide)s, like poly(R-glutamic) acid,15 revealing that the insertion, in the R-carbonyl position, of two CH2 groups does not introduce significant changes in the helical fold since the dihedral angles associated with them adopt a trans conformation. These conformational trends allow one to understand the resemblance in both size and shape between the helices of PGDGA and poly(R-glutamic) acid. The rmsd from the model structure of PGDGA20 at 350 K increases considerably with respect to that of the same compound at 300 K (Figure 2a). Thus, for the first 400 ps, the model structure remains stable, retaining more than 60% of the initial hydrogen bonds (Figure 3a). Indeed, the most remarkable feature along this period of time is the fraying of the N-terminus side (Figure 4b). However, a very fast and complete conformational transition toward an unfolded conformation occurs after 400 ps (Figure 4b). Thus, the rmsd increases from 2.5 Å to more than 7 Å, and the fraction of hydrogen bonds decreases from about 60% to 10% in only 100 ps. This random coil remains for the last 500 ps of simulation, even although along this period nuclei of helix occasionally appeared but soon disappeared. Figure 5b illustrates the unfolding process as well as the helix nucleus formed at 600 and 800 ps.

MD simulations of PADGA10 at both 300 and 350 K provided very similar results. Thus, in both cases the initial helical conformation quickly transforms into a random coil state as is indicated by the evolution of both the rmsd (Figure 2b) and the fraction of i to i + 3 hydrogen bonds (Figure 3b). Furthermore, inspection of the spatiotemporal evolutions of such hydrogen bonds in the first 200 ps reveals that the unfolding process starts at the C-terminus side (Figure 4c,d). These results indicate that the stability of the helical conformation of PGDGA is strongly influenced by the length of the polypeptide chain. Thus, chains containing only 10 residues are not able to retain a regular secondary structure rapidly evolving toward a random coillike state. Energy Analysis. The results obtained for PGDGA20 at 300 and 350 K indicate that the helix is more stable at room temperature. To explain this feature, we have investigated the helix disruption from an energetic point of view using a procedure similar to that employed by Takano and coworkers16 for polyalanine. For this purpose, we computed the peptide-peptide (p-p), peptide-water (p-w), and water-water (w-w) interactions at the trajectories. We only analyzed the first 1000 ps of the total trajectory obtained at 300 K in order to have a better comparison between the results provided by the two temperatures. However, we cannot expect important changes in the analysis since the helix remain stable during the last 500 ps (Figures 2-4). Figures 7a and 8a show the parametric plots between the p-p intramolecular energies and the fraction of the native i to i + 3 hydrogen bonds in the PGDGA20 chain during the simulations at 300 and 350 K, respectively. The p-p potential energy decreases as more initial hydrogen bonds are retained indicating that this intramolecular energy contribution is more favorable for the helix than for the random coil. Figures 7b and 8b show the parametric plots between the p-w energies and the ratio of native hydrogen bonds during

Biomacromolecules, Vol. 2, No. 3, 2001 655

Poly(γ-glutamic acid) in Aqueous Solution

Figure 7. Parametric plots between the (a) peptide-peptide (p-p), (b) peptide-water (p-w), and (c) water-water (w-w) energies and the ratio of i to i + 3 hydrogen bonds in the PGDGA20 chain during the simulation at 300 K.

the two simulations. As expected the fraction of formed hydrogen bonds has a large influence on the strength of such interactions. The p-w energies are considerably more favorable for the random coil than for the helix indicating that the solvation is worse for the latter conformation than for the former one. This is also reflected by the variation of the solvent-exposed surface area of the polar atoms, i.e., N, H(-N), and O atoms, which increases with the unfolding of the helix. Thus, this area has been measured to be about 973 ( 41 and 1242 ( 50 Å2 for the helix and coil of PGDGA20, respectively. The greater solvent-accessibility of the polar groups in the random coil state allows one to increase the number of interactions with the surrounding water molecules. Finally, the w-w energies (Figure 7c and 8c) slowly increase when the fraction of formed hydrogen bonds decreases. This behavior is consistent with the variation of the p-w energy. Thus, the amount of w-w interactions decreases when the number of p-w interactions increases. The sum of the p-p, p-w, and w-w energy changes associated with the helix disruption may be connected with the potential energy change in unfolding the helix. These values were obtained by applying a linear correlation to the parametric plots of Figure 8. The resulting fitted lines and the energies associated with the conformational change are displayed in Table 1. As it can be seen there is a good correlation between the ratio of formed hydrogen bonds and the different energy terms, the Pearson regression coefficients (r) being about 0.7-0.9. This points out the crucial role of the i to i + 3 hydrogen bonds in our simulations since they

Figure 8. Parametric plots between the (a) peptide-peptide (p-p), (b) peptide-water (p-w), and (c) water-water (w-w) energies and the ratio of i to i + 3 hydrogen bonds in the PGDGA20 chain during the simulation at 350 K. Table 1. Energy Changes (in kcal/mol) from the Helix to the Coil State for PGDGA20. energy terma

fitted lineb

rc

energy changed

p-p p-w w-w system

y ) -1221.02 - 2.23x y ) -886.62 + 3.39x y ) -7650.20 - 1.21x

0.9 0.9 0.7

223 -339 121 5

a p-p, p-w, w-w, and system refer to the peptide-peptide, peptidewater, water-water, and whole system interactions, respectively. b The fitted lines were obtained by applying a linear correlation to the parametric plots between the energies and the native hydrogen bond ratio as shown in Figure 8; x and y refer to the ratio of hydrogen bonds and the energy, respectively. c r refers to the Pearson regression coefficient. d The energy changes were obtained by applying the fitted lines to the helix (x ) 1) and coil (x ) 0) states.

affect not only to the p-p energy term but also to the p-w and w-w ones. A linear correlation was also found for polyalanine.16 It is worth noting that the solvation potential, which is estimated from the sum of the p-w and w-w energy changes, favors the random coil by about 218 kcal/mol. On the other hand, the p-p potential energy change predicted from the equation of Table 1 indicates that the helix conformation is 223 kcal/mol more stable than the coil state. According to these values, the energy change that accompanied the helix-coil transition in the MD at 350 K was about 5 kcal/mol, favoring the helix. It should be mentioned that these values should be considered only from a qualitative point of view since the conformational space of the random coil has not been exhaustively explored. However, this complex task is beyond the scope of the present work.

656

Biomacromolecules, Vol. 2, No. 3, 2001

Figure 9. Temporal evolution of the distribution for the torsional angle χ1 in the simulation of PDGDA20 at 350 K. The population analyses have been performed each 100 ps. The four categories considered, in the order displayed in the figure from left to right, are trans, gauche+, gauche- and the remaining conformers.

It would be interesting to compare the energetic results of PGDGA with those obtained for other helical peptides. For a 15-residue polyalanine in aqueous solution, the energy is lower in the R-helix state than in the random coil state by 20.4 kcal/mol.16 This is in agreement with the values obtained by the calorimetric experiment of Scholtz et al.17 (13-19 kcal/mol). Accordingly, the R-helix of polyalanine seems to be more stable than the helix of PGDGA in peptides constituted by 15 or 20 residues. This could be attributed to the lower density of amide‚‚‚amide hydrogen bonds in the latter. Thus, the two additional backbone carbon atoms of the PGDGA residues increase considerably the distance between consecutive hydrogen bonds. On the other hand, γ-peptides constituted by amino acids without polar side groups have been shown to form helical secondary structures more stable than R-peptides in pyridine and methanol solutions.18 However, the helices of such γ-peptides are substantially different from that of PGDGA since they contain 14-membered ring hydrogen bonds set between the NH of the amide group i and the CO of the amide group i + 2. This structural behavior is consistent with that of helical poly(β-peptide)s. Thus, the helix conformation of the poly(β-peptides)s constituted by amino acids containing polar groups in the side chain, i.e. poly(R-alkyl-β-L-aspartate)s, is less stable than the R-helix of poly(R-peptide)s,19 but the opposite situation appears when the chain is constituted by β-amino acids without polar side groups.20 Side Chain Conformation of PGDGA. To get a deeper insight into the structure of PGDGA, the conformational preferences of the side chain torsional angle χ1 were analyzed. For this purpose population analyses were performed for the four simulations considered in this work, the conformations being grouped in four categories: gauche+ (from 30 to 90°), trans (from +150 to -150°), gauche- (from -90 to -30°) and the remaining conformers. The temporal evolution of the torsional angle distribution for the simulation of PDGDA20 at 350 K is displayed in Figure 9, which was obtained by analyzing the structures saved each 100 ps. The results provided by the simulations of PGDGA20 at 300 K and PGDGA10 at both 300 and 350 K (data not shown) were similar to those displayed in Figure 9, no significant difference being found.

Zanuy and Alema´ n

Figure 10. Number of interactions between the backbone amide group and the side chain oxygen atoms found in the simulations of PGDGA20 (solid line) and PGDGA10 (dashed line) at 300 K.

As it can be seen, the gauche- conformation is the preferred conformation for the PDGDA side chain. Thus, this side chain conformation is clearly the most populated one independently of the backbone arrangement. Furthermore, it becomes the most populated in the first steps of the simulation since the initial trans conformation is rapidly changed. The population of gauche+ conformation increases with the unfolding of the helix, being of about 20% in the random coil state. On the other hand, the population of trans is almost negligible throughout all the simulation. Finally, the population of cis (from -30 to +30°), skew+ (from 90 to 150°) and skew- (from -150 to -90°), which are grouped under the category of remaining conformers, is nonnegligible, i.e. about 20-30%. It should be noted that the gauche- conformation was predicted to be one of the most stable conformations for the χ1 dihedral angle of PGDGA.9 This rotamer allows an intraresidue interaction between the hydrogen atom of the amide group and the alkoxy oxygen atom of the acid side group. On the other hand, it was demonstrated that the helices adopted by some polypeptides composed of R- and β-amino acids bearing polar side groups are partially stabilized by intramolecular interactions between the hydrogen amide atom and the oxygen atoms of the side chain.21,22 This specific interaction is also present in PGDGA not only in the helical conformation but also in the coil state. Figure 10 shows the number of interactions between the backbone amide groups and the oxygen atoms of the side chain for PGDGA20 and PGDGA10 at 300 K. Similar results were obtained from the simulations at 350 K. With this work, we have displayed the effect of both the temperature and length of the peptide on the dynamics of the helical conformation of PGDGA. We have shown that the helix is not stable for a 10-residue peptide independently of the temperature. However, the temperature plays a crucial role on the dynamics of the 20-residue peptide. At room temperature, PGDGA20 retains the helical conformation throughout the simulation, while at 350 K, the helix spontaneously evolves toward the random coil state. A potential energy analysis indicates that the helix state is less stable than the random coil state, even although the latter is better solvated than the former.

Poly(γ-glutamic acid) in Aqueous Solution

Acknowledgment. This work was supported by DGICYT with Grant No. PB96-0490. D.Z. acknowledges the support of the Ministry of Education of Spain for the award of a scholarship. The authors are indebted to the Centre de Supercomputacio´ de Catalunya (CESCA) for computational facilities and to Prof. S. Mun˜oz-Guerra for a helpful discussion. References and Notes (1) Thorne, C. B.; Go´mez, C. G.; Noyes, H. E.; Housewright, R. D. J. Bacteriol. 1954, 68, 307. (2) Ko, Y.; Gross, R. A. Biotechnol. Bioeng. 1998, 57, 430. (3) Cromwick, A.-M.; Gross, R. A. Int. J. Biol. Macromol. 1985, 49, 846. (4) Pe´rez-Camero, G.; Congregado, F.; Bou, J. J.; Mun˜oz-Guerra, S. Biotechnol. Bioeng. 1999, 63, 110. (5) Gardner, J. M.; Troy, F. A. J. Biol. Chem. 1979, 254, 6262. (6) Troy, F. A.; Frerman, F. E.; Health, E. C. J. Biol. Chem. 1971, 246, 118. (7) Rydon, H. N. J. Chem. Soc. 1964, 1328. (8) He, L. M.; Neu, M. P.; Vanderberg, L. A. EnViron. Sci. Technol. 2000, 34, 1694. (9) Zanuy, D.; Alema´n, C.; Mun˜oz-Guerra, S. Int. J. Biol. Macromol. 1998, 23, 175. (10) Zanuy, D.; Alema´n, C.; Mun˜oz-Guerra, S. Macromol. Theory Simul. 2000, 9, 543.

Biomacromolecules, Vol. 2, No. 3, 2001 657 (11) Weiner, F. J.; Kollman, P. A.; Nguyen, D. T.; Case, D. A. J. Comput. Chem. 1986, 7, 230. (12) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. J. Chem. Phys. 1983, 79, 926. (13) Ryckaert, J. P.; Ciccoti, G.; Berendsen, H. J. C. J. Comput. Phys. 1977, 23, 327. (14) Pearlman, D. A.; Case, D. A.; Caldwell, J. W.; Ross, W. S.; Cheatham, T. E. III; Ferguson, D. M.; Seibel, G. L.; Singh, U. C.; Weiner, P. K.; Kollman, P. A. AMBER 4.1, University of California, San Francisco, 1995. (15) Zanuy, D.; Alema´n, C. Biopolymers 1999, 49, 497. (16) Takano, M.; Yamato, T.; Higo, J.; Suyama, A.; Nagayama, K. J. Am. Chem. Soc. 1999, 121, 605. (17) Scholtz, J. M.; Marquesse, S.; Baldwin, R. L.; York, E. J.; Stewart, J. M.; Santoro, M.; Bolen, D. W. Proc. Natl. Acad. Sci. U.S.A. 1991, 88, 2854. (18) Hintermann, T.; Gademann, K.; Jaun, B.; Seebach, D. HelV. Chim. Acta 1998, 81, 983. (19) Martı´nez de Ilarduya, A.; Alema´n, C.; Garcı´a-Alvarez, M.; Lo´pezCarrasquero, F.; Mun˜oz-Guerra, S. Macromolecules 1999, 32, 3257. (20) Seebach, D.; Gademann, K.; Schreiber, J. V.; Matthews, J. L.; Hintermann, T.; Jaun, B.; Oberer, L.; Hommel, U.; Widmer, H. HelV. Chim. Acta 1997, 80, 2033. (21) Leo´n, S.; Alema´n, C.; Mun˜oz-Guerra, S. Macromolecules 1997, 30, 6662. (22) Alema´n, C.; Navas, J. J.; Mun˜oz-Guerra, S. J. Phys. Chem. 1995, 99, 17653.

BM000122J